Financial Fraud Detection Approach Based on Firefly Optimization Algorithm and Support Vector Machine. (17th June 2022)

• Record Type: Journal Article
• Authors: Singh, Ajeet; Jain, Anurag; Biable, Seblewongel Esseynew
• Other Names: Ramachandran Manikandan Academic Editor.
• Is Part Of: Applied computational intelligence and soft computing. Volume 2022(2022)
• Publication Date: 2022-06-17
• Subjects: Computational intelligence -- Periodicals; Soft computing -- Periodicals; 006.305
• DOI: 10.1155/2022/1468015
• Languages: English
• ISSNs: 1687-9724

Abstract

Abstract : The usage of credit cards is increasing daily for online transactions to buy and sell goods, and this has also increased the frequency of online credit card fraud. Credit card fraud has become a serious issue for financial institutions over the last decades. Recent research has developed a machine learning (ML)-based credit card fraud transaction system, but due to the high dimensionality of the feature vector and the issue of class imbalance in any credit card dataset, there is a need to adopt optimization techniques. In this paper, a new methodology has been proposed for detecting credit card fraud (financial fraud) that is a hybridization of the firefly bio-inspired optimization algorithm and a support vector machine (called FFSVM), which comprises two sequential levels. In the first level, the firefly algorithm (FFA) and the CfsSubsetEval feature section method have been applied to optimize the subset of features, while in the second level, the support vector machine classifier has been used to build the training model for the detection of credit card fraud cases. Furthermore, a comparative study has been performed between the proposed approach and the existing techniques. The proposed approach has achieved an accuracy of 85.65% and successfully classified 591 transactions, which is far better than the existing techniques. The proposed approach has enhanced classification accuracy, reduced incorrect classification of credit card transactions, and reduced misclassification costs. The evaluation results show that the proposed FFSVM method outperforms other nonoptimization machine learning techniques.